References & Citations
Statistics > Methodology
Title: A Unified Computational and Theoretical Framework for High-Dimensional Bayesian Additive Models
(Submitted on 14 Jul 2020 (v1), revised 21 Sep 2020 (this version, v4), latest version 29 Jun 2023 (v6))
Abstract: We introduce a general framework for estimation and variable selection in high-dimensional Bayesian generalized additive models where the response belongs to the overdispersed exponential family. Our framework subsumes popular models such as binomial regression, Poisson regression, Gaussian regression, and negative binomial regression, and encompasses both canonical and non-canonical link functions. Our method can be implemented with a highly efficient EM algorithm, allowing us to rapidly attain estimates of the significant functions while thresholding out insignificant ones. Under mild regularity conditions, we establish posterior contraction rates and model selection consistency when the number of covariates grows at nearly exponential rate with sample size. We illustrate our method on both synthetic and real data sets.
Submission history
From: Ray Bai [view email][v1] Tue, 14 Jul 2020 13:29:13 GMT (54kb,D)
[v2] Sun, 26 Jul 2020 19:57:31 GMT (55kb,D)
[v3] Wed, 29 Jul 2020 06:13:38 GMT (55kb,D)
[v4] Mon, 21 Sep 2020 16:57:47 GMT (57kb,D)
[v5] Sat, 5 Jun 2021 15:22:40 GMT (101kb,D)
[v6] Thu, 29 Jun 2023 03:15:45 GMT (43kb)
Link back to: arXiv, form interface, contact.